oss-code-analysis
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted content from external GitHub repositories, which is an inherent part of its research functionality. This creates a theoretical surface for indirect prompt injection if a repository contains malicious instructions in its README or source code.
- Ingestion points: The skill fetches file contents and metadata from github.com and raw.githubusercontent.com as specified in the SKILL.md procedure.
- Boundary markers: The skill includes a dedicated Guardrails section that specifically commands the agent to treat all fetched file contents as untrusted external data and to never interpret or execute embedded instructions or agent directives.
- Capability inventory: The skill utilizes web search and fetching capabilities for reading analysis material; it does not have tools for shell command execution, file system modification, or data exfiltration.
- Sanitization: The instructions mandate that repository content is used exclusively for analysis and must not alter the skill's procedure, output structure, or tool usage.
Audit Metadata